D617 - A second batch, and the pipeline that was built for it

Status: measured Date: 2026-09-08

What arrived

Six files, timestamped 20260908-143022 - an hour after the batch D609 was built around, and replacing it in the same directory. Three contexts again, and twelve sections nobody had captured before:

015-sync/mutex-unlock-unheld        102-net/listener
018-relational/allocations-do-not-overlap   102-net/nonblocking-option
018-relational/handle-fits-its-out-parameter 102-net/recv-would-block
080-video/attribute-block           102-net/resolve
101-input-ext/keyboard-held         102-net/sockaddr-bind
101-input-ext/mouse-moving          107-videodec/create-present
101-input-ext/reachability          108-audiodec/decode-present
110-modules/symbol                  135-sysctl/osrelease

Nothing had to be built to read them. orbistoun-gen measurements picked all three up from the directory it already reads, names --from-report picked up the export table, and the coverage gate failed with a list. That is what the morning's work was for, and this is the first time it has been asked to do it unattended.

	after D609	now
distinct measurements	374	429
constant across every run	262	304

Forty-two new constants, and nothing previously claimed was demoted - nothing_claims_a_measurement_that_cannot_be_claimed passed on the first attempt, which it did not last time.

The nine worth having

Thirty-three of the forty-two are console addresses, that machine's handle numbering, or how long the probe waited for an operator. Opaque, on the same terms as before.

The other nine are a subsystem orbistoun has not started. None of sceNetSocket, sceNetBind, sceNetListen, sceNetAccept, sceNetConnect, sceNetRecv, sceNetSend or sceNetSetsockopt is declared here, and the console answered for all of them:

sceNetBind        bound          0x0
sceNetListen      listening      0x0
sceNetSetsockopt  0x1200-accepted 0x0
sceNetRecv        connected-return  0x80410123
sceNetRecv        listener-return   0x80410139

That last pair is the reason to record a measurement before writing the function. A socket with nothing to read answers one code when it is connected and a different one when it is listening - a distinction a reimplementation flattens without noticing, because both read as "would block". And the 0x8041 prefix is libSceNet's own error space, not the 0x8002 kernel one that every existing vendor code in this project uses.

Written into libSceNet.toml as measured, citing the check, for a function that does not exist yet. That is the inversion this whole apparatus is for: the oracle arrives first and the implementation is written against it, rather than being written against a guess and corrected when a title disagrees.

The export table did not move

2,443 exports in this batch, 215 of them unnamed - the same counts as the batch an hour earlier, and the affix search found nothing new. Whether the addresses also match is unverified: the previous batch was replaced rather than archived, so there is nothing left to compare against.

Worth noting as a limit of the arrangement rather than a result. A report directory that is overwritten cannot answer "did this change", which is the question two captures exist to answer.